JPH07193661A - Telephone marketing complex management system - Google Patents

Abstract

PURPOSE: To improve the management system capability of a telephone marketing performance by automatically setting a threshold level by a telephone marketer. CONSTITUTION: A probability that a performance object(PO) is satisfied by an agent which controls an automatic call distributor of a telephone marketing complex is calculated with respect to each sub-interval of a prescribed service interval. For the purpose of meeting the PO, this probability is compared with a threshold level(TL). The threshold level used to determine whether a call should be shifted from a prescribed automatic call distributor(ACD) or should be received is automatically set by selection of the telephone marketer related to a proportion of false alarm or false reception allowable in the telephone marketer. Thus, the capability of the telephone marketing performance management system is considerably improved.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates to a telemarketing complex performance management system. More specifically, the present invention changes the number of employees at a particular traffic center to determine when to reroute traffic to level service levels among automatic call distributors, or to the particular traffic center where the call is received. The present invention relates to a telephone marketing complex performance management system in which a threshold level is set in order to do so.

[0002]

In the Telephone Marketing Complex Performance Management System (TCPMS), the Telephone Marketing Complex compiles and processes data obtained from customer premises and public switched telecommunications network sources. System, which allows service levels to be leveled among telecommunications centers (TCs) within the complex.

Generally, a telephone marketing complex consists of multiple automatic call distributors (ACDs) and multiple data centers. The data center can be selectively connected to the TC on which the telephone marketer receives the call. This system generally includes the following elements: (1) Telephone in all places
An interface to the network that reports the origin and destination of each of the multiple calls received by the marketer, (2) an interface to the network that allows the telephone marketer to change the call processing logic, and (3) other systems Traffic Control Center Processor (TCCP) that accepts input data from all elements of the.

The system accumulates data from each ACD and telephone marketer database (located at a data center (DC)) in the TCCP, while at the same time originating and terminating multiple calls received at each ACD. Both are monitored and, with it, a call processing logic change is initiated. The system parses the data and generates signals to implement the call routing changes necessary to level the service levels across ACDs or the change in staffing at the particular TC at which the call is received. This process is performed in real time and is repeated at the desired frequency.

US Pat. No. 5,164,983 discloses a Telephone Marketing Complex Performance Management System (TCPMS). With this system, telephone marketing customers must select a performance objective that they want to be met during a selected service interval.

In this system, the performance goal is the average response speed (ASA) by personnel at a given ACD governing the operations being performed by the telephone marketing complex. The reactive process in the system determines, for each subinterval within the service interval, the probability of meeting the performance objective (P _s ) at the ACD. This probability is then compared in reaction processing against two threshold levels, T _lo and T _hi .

These two threshold levels are also selected by telephone marketing customers. If the probability P _s is less than or equal to T _lo , an alarm condition is raised. This alert condition indicates that the ACD under consideration shifts the call to increase the probability of meeting performance goals during the service interval. On the other hand, the probability P _s is T _hi
If so, there is a receptive state. This acceptance state indicates that the ACD under consideration is easily loaded and can receive traffic from other ACDs in the system.

Of course, if the probability is between two threshold levels, there is neither an alarm condition nor an acceptance condition, and the reactive process does not indicate a change in the traffic routing pattern.

In the system in US Pat. No. 5,164,983, telephone marketing customers
Different values of T _hi and T _lo can be set for each sub-interval within the service interval at the terminal wishing to achieve the selected performance target value.
By repeating trial and error, the customer can reach a set of threshold values. This set of threshold values provides the desired traffic exchange throughout many of the fluctuations in incoming traffic caused by seasonal, daily and monthly changes of telephone marketer calling customers.

However, conventional telephone marketing complex performance management systems have not been fully satisfactory. In particular, the selection of a performance target and the setting and changing of T _hi and T _lo associated therewith have a serious difficulty in that they cannot be realized without repeating trial and error.

[0011]

Accordingly, it is an object of the present invention to provide an improved telephone marketing complex performance management system.

[0012]

SUMMARY OF THE INVENTION An improved telephone marketing complex performance management system in accordance with the present invention determines whether a call should be shifted from or accepted by a given automatic call distributor. The threshold level used is set automatically by the telephone marketer's selection of the false alarms or rate of false acceptances that the telephone marketer can accept.

In particular, the calculated value for the probability of successfully achieving the performance objective P _s is saved for each subinterval within the service interval. At the end of the service interval, an indication of success or failure to achieve the performance goal is entered into a memory table of probabilities for each subinterval within the service interval. Each subinterval has its own memory table, and success or failure indications are entered into a memory table within the probability range that contains the calculated probability of success for this subinterval.

A history of successes and failures is maintained in each subinterval table for a given number of preceding intervals. The table for subintervals is then used to calculate _Tlo based on the percentage of false alarms that the telephone marketer can accept. An alert is generated when the system predicts that the agent in charge of this ACD will not be able to meet its performance goals unless the call is shifted from the ACD.

Similarly, the calculation of T _hi is also performed using a table for the same subinterval based on the percentage of false acceptance signals that the telephone marketer can accept. If the ACD can accept the call,
An acceptance signal is generated when the system predicts that the agent on the CD can still meet the performance goal.

The calculation of the call load rate, which must be done during a given subinterval to achieve a desired ASA-like performance goal, requires a number of available queue positions. The average line estimation coefficient is calculated during each service interval. This line estimation factor gives an estimate of the number of lines used by the application under consideration when multiplied by the number of agents. This line estimation factor minimizes the mean squared error (MSE) between the estimated call load as determined by the line estimate factor and the actual call load experienced over a given number of conventional subintervals. Is easily calculated by

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be specifically described below with reference to the drawings.

As pointed out in US Pat. No. 5,164,983, the telephone marketing complex 101 of FIG. 1 of the present invention includes a plurality of telecommunications centers (T).
C). This telecommunications center (TC) can be located in different geographical locations. Generally, each TC
Is at least one automatic call distributor (AC
D) and one or more "split / gate".
The "split / gate" receives information from the attached ACD at the TC.

"Split" or "gate" means a group of attendant stations (sometimes referred to as agent terminals). Each attendant station comprises input / output terminals and / or other display means. The display means allow the attendant at the TC to handle the particular call received by the ACD. A data center (DC) is associated with each TC. A data center (DC) typically contains a database containing customer profiles and other information needed to handle incoming calls.

In order to manage and control the telephone marketing complex and to be able to efficiently process the data inside these, the telephone marketing complex 101 has the data shown in FIG. And a data acquisition and control (DAC) module 102. The DAC module 102 includes an interconnect carrier (I
Data from the XC) network 103 is also provided.
A very complete description of the data collected by the DAC 102 is shown in FIG. 3 of US Pat. No. 5,164,983 and is set forth in the relevant portion of the patent specification in connection with this figure.

In short, the data is the AC at each TCD.
Related to D-Performance and Telephone Marketer Composition. All of this data is stored in memory as shown by live data and configuration information storage block 104 in FIG. This block 104 is described in US Pat.
It corresponds to the block 400 and the block 410 of FIG. 4 of 64983 specification.

The stored live data and configuration information is available to the module 105 of FIG. Module 105 monitors the actual call load and makes an estimate of the required call load needed to meet the target during the current service interval. Using queuing theory,
Describe the average answer speed (ASA) equation in terms of the average staffing level for the current service interval (n), the average service speed (μ), the call load speed (λ) and the number of queue positions (N). You can

Crane Lusec & Co.
See page 104 of the text entitled "Theory of Queuing in OR" by E. Page, published by (Crane Russek & Company) in 1972.

The number of queue positions (N) can be estimated by using the following equation: N = (T-1) · n (where T is the line estimation coefficient and n is the average staffing level.) Use to use the queuing relationship for ASA and to meet performance goals A
Knowing the SA allows us to derive the call load rate needed to meet the service objective (λ _req ).
This call load rate can then be used to derive the number of calls (R).

The number of calls must be provided for the remaining service interval if the performance goal is met with acceptable certainty by using the following relationship: λ _req = (NCO _t −NCA _t + R) / (t + τ) (where NCO _t is the number of calls provided from the start of the service interval to the current subinterval having an elapsed time of _t, NCA _{t (t} is the number of calls abandoned by the current subinterval, and τ is the time remaining in the service interval.)

Given the prior distribution of call load rates for applications in TC and the required number of calls (R) for the remaining service interval (τ), the remaining interval is estimated according to the Poisson distribution, based on an estimate of arriving calls. A predictive distribution is established for the number of calls that arrive inside. Then, using the negative binomial predictive distribution, module 106 of FIG.
Calculate the probability of receiving a book call.

Et Reifer, published in 1961 by the Research Department of Harvard Business School
21 of the text entitled "Applied Statistical Decision Theory" by H. Raiffa and R. Schlaifer.
See pages 6,236, and 284. At most R
The probability of receiving a book call is equal to the probability of meeting the performance goal given to the current ASA.

The probability of meeting the performance goal obtained from module 106 is then compared in module 107 with the alarm and acceptance thresholds, T _lo and T _hi , respectively. Ps calculated in module 106
Is less than or equal to the threshold T _lo , module 10
7 generates an alarm condition which causes the module 108 to determine the number of calls to shift from the ACD under consideration in order to achieve the performance goal during the service interval.

Similarly, if P _s is greater than or equal to the threshold T _hi , module 107 will generate an accepting state, which causes module 109 to determine the number of calls that can be accepted from another ACD.

The module thus far described is functionally identical to the module described in US Pat. No. 5,164,983. In the embodiment of the present invention, a service interval of 30 minutes is selected and each subinterval is 3 minutes. It has been discovered that subintervals as long as 3 minutes smooth out the instantaneous variations in _Ps values that may be present in subintervals as short as 40 seconds.

Therefore, in US Pat. No. 5,164,983, the alarm and acceptance thresholds had to be predetermined by the telephone marketing customer and selected for each 3 minute subinterval. According to the present invention, the alert and acceptance threshold levels, T _lo and T _hi, are calculated by the modules described below according to the customer's selection of the number of false alerts and false acceptances a telephone marketing customer can accept.

After determining the number of calls to shift (if any) or the number of calls to accept (if any), processing proceeds to module 201 of FIG. In this module 201, line estimation coefficients are determined. In conventional systems, this factor was selected by reviewing historical data and has a value of approximately 1.45. According to the invention, this factor is adjusted during each subinterval by the method described in more detail below in connection with FIG.

After determining the line estimation coefficients, processing proceeds to decision module 202. The decision module 202 determines whether the current subinterval is less than the last subinterval of the service interval. If it is less than the last subinterval, the process does not move the last subinterval and the process proceeds to module 203. Module 203 saves the P _s value for this subinterval, and the process continues with alternative process 302 of FIG.

This alternative process is described in US Pat. No. 5,164,984.
It is the same as the alternative process relating to FIG. In essence, the alternative process changes the call processing logic that manages the distribution of the call load between ACDs and / or changes the worker level in the TC to be implemented.
Decide on either. As in US Pat. No. 5,164,983, all modification recommendations generated by the alternative process are forwarded as input to the external system interface module 303 of FIG.

This external system interface module 303 provides the display at the telephone marketing complex manager's workstation. As in U.S. Pat. No. 5,164,983, the external system interface module 303 also provides output to the IXC interface 103 and / or predictor of the workforce as needed.

When the last subinterval is detected by the decision module 202 of FIG.
Proceed to 04, and first, whether the telephone marketing complex succeeded in achieving the performance goal,
That is, the average response speed (A
SA) is below the performance target. This failure or success indication is then placed in the memory table for each subinterval at the position in the table that corresponds to the _Ps value obtained during this subinterval.

A histogram display of such a table is shown in FIG. The table shown in FIG. 5 is for a 3 minute subinterval of 360 to 540 seconds from the start of the service interval. A similar table can be provided for each subinterval. "success"
And the number “1” in the column below the section with the indication of “failure”
Shows one case where P _{s for} this subinterval had a probability within the range of probabilities shown for the column containing this number.

In actual processing, each digit "1" may also have an indication as to the date of origination. This allows the module 301 to periodically erase the memory that stores information that is older than one week.

Thereafter, processing proceeds to module 205 to calculate the T _lo value based on the information stored by module 204. As shown in the histogram of FIG. 5, there are 6 subintervals under consideration with P _s values less than 20%, nevertheless the system met the performance goals well.

The total number of cases where the P _s value is less than 20% is 2
Since there were 0 cases, the probability of generating a false alarm is 30% (6
/20=0.3). Therefore, if the customer requests a false alarm occurrence probability of 30% or less, the threshold level T _lo can be set to 20%. Similar calculations can be performed in module 205 to determine a particular threshold level that can meet another customer requirement for false alarm probability.

Processing then proceeds to module 206 to calculate the T _hi value based on the information stored by module 204. As shown in the histogram of FIG. 5, there are 3 sub-intervals under consideration with P _s values above 70%, but the system nevertheless failed to meet the performance goals.

Since there were a total of 19 cases with more than 70%, the probability of generating a false alarm was 15.8% (3/19 =
0.158). Therefore, if the customer requests that the probability of false acceptance be less than 16%, for example,
Module 206 sets the acceptance threshold level T _hi to 70%. As before, to meet another value for the probability of false acceptance, module 206 can perform a similar calculation to determine another T _hi value.

The process then proceeds to module 301 in FIG. In module 301, modules 205 and 2
The T _lo value and T _hi value determined in 06 are saved. As a result, the new value can be used by module 107 of FIG. 1 during the next service interval. Then, the threshold values of T _lo and T _hi are periodically re-evaluated by the method of the present invention, and consequently updated by changing the conditions to meet the customer selection probability of generating false alarms and false acceptances. .

Finally, processing proceeds from module 301 to alternative processing 302 and the method as described in US Pat. No. 5,164,983 is continued.

Module 20 for determining line estimation coefficients
An extension of 1 is shown in FIG. Initial values of the line estimation coefficient (T) and the parameter (k) are retrieved from the memory in the module 401 of FIG. At the start of the process, the value of T is set to 1.45 and the value of the parameter k is set to 1.00 each time the module 401 starts. afterwards,
In module 402, for all history records for the current subinterval, the actual call load and
Compute the difference in estimated call load based on the line estimation coefficients of T, T + k and T−k.

Processing then proceeds to module 403 of FIG. In module 403, T, T + k and T-
Calculate the mean squared error (MSE) for the difference obtained using each line estimation coefficient of k. As a result, three distinct mean squared errors are generated for the subintervals. Each error corresponds to each line estimation coefficient under consideration.

Then, in module 404, the value of (T + k) or (T-k) for the line estimation coefficient is T.
Consider whether you have generated a mean squared error that is lower than the value of. If so, the process moves to decision module 407. The value of k is subtracted from T in module 408 or added to T in module 409.
(In either case, it produces a low mean squared error.) The process then returns to module 402.

Thus, the value of the line estimation coefficient is
We move towards values that tend to achieve lower mean squared errors in the new calculation of the difference. On the other hand, if the lowest mean squared error was generated at module 403 by the current T value, then decision module 404 moves processing to module 405. In module 405, the value of k is halved. If the new value of k is greater than or equal to 0.000007629, the decision module 406 returns the process to the calculation module 402 along with the reduced value of k.

This iterative process of calculating the mean square error is
If the value of k is less than the value indicated in module 406, it is stopped, then the process moves to module 412. At module 412, the value of T for the current subinterval is stored in memory.

Finally, the average value of T is calculated in module 410 using the T value for each subinterval of the service interval. The average value of T is then stored in memory by module 411 and the processing associated with module 201 of FIG. 2 is stopped.

The sequences for determining the threshold levels of Tlo and Thi can be interchanged with each other without compromising the effectiveness of the processing method of the present invention. Furthermore, the method of the present invention can also be used to optimize system performance for requirements other than ASA. The range of probability values used in the histogram of FIG. 5 can be either larger or smaller than the 5% range used in the previous example.

[0052]

As explained above, in the prior art US Pat. No. 5,164,983 the alarm and acceptance thresholds must be preselected by the telephone marketing customer. The determination of alerts and acceptance thresholds has been done through repeated trial and error. However, in accordance with the present invention, the threshold level used to determine whether a call should be shifted or accepted by a given automatic call distributor is a false level that telephone marketers can accept. Set automatically by telephone marketer's choice regarding alarm or false acceptance rate. As a result, the capabilities of the telephone marketing performance management system are significantly improved.

[Brief description of drawings]

FIG. 1 is a flow chart of a portion of the processing performed by the method of the present invention.

FIG. 2 is a flowchart of a portion of the processing performed by the method of the present invention.

FIG. 3 is a flowchart of a portion of the processing performed by the method of the present invention.

FIG. 4 is a flow diagram of the process used to determine line estimation coefficients.

FIG. 5 is a table histogram of successes and failures during a single subinterval, useful in explaining the method of the present invention.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Robert A. Marumi United States, 07701 New Jersey, Red Bank, West Wilson Circle 90 (72) Inventor Bernadette Em. Strout United States, 80304 Colorado, Boulder, Northridge Court 1363

Claims (8)

[Claims] 1. The probability that a predetermined performance goal will be achieved by an automatic call distributor within the telephone marketing complex at the end of the service interval in the telephone marketing complex management system is calculated for each subinterval of the service interval, The management system has means for comparing the calculated probabilities with a threshold level for shifting calls and a threshold level for accepting calls, and (1) successful achievement of a performance goal for each subinterval within the service interval. And (2) accepting the call and the threshold level for shifting the call for each subinterval within the service interval in response to the record keeping means. Telemarketing complex management system characterized by having a combination comprising a means for generating the threshold level that. 2. The means for responding to the record holding means comprises:
Means for generating a threshold level for shifting a call based on a preselection probability for generating a false alarm to shift and (b) a threshold level for accepting a call based on a preselection probability for generating a false signal to accept The telephone marketing complex management system of claim 1 having both means for generating. 3. A telephone marketing complex management system having a plurality of automatic call distributors (ACDs) for receiving calls placed in the telephone marketing complex, and means for obtaining and storing data from said ACD. Responsive to the acquisition and storage means, means for calculating a probability of achieving a performance goal at the end of a service interval for each of the ACDs in the telephone marketing complex, and a threshold level at which the calculated probability shifts a call. An alert condition if less than or equal to, and an accept condition if the calculated probability is greater than or equal to a threshold level for accepting a call, during each subinterval within the service interval. And the means for generating And,
Means for determining the number of calls to be shifted from each of the ACDs, means for determining the number of calls to be accepted in each of the ACDs in response to the generating means, and the number of calls to be shifted and the number of calls to be accepted In response, the AC
In a telephone marketing complex management system comprising means for determining a change in call load distribution among Ds and means for displaying a change in call load distribution among the ACDs, the change determining means is for each subinterval. A record of the calculated probabilities and a means for elaborating a record of whether the system has successfully or failed to achieve the performance goal within the service interval, and, in response to the record keeping means, for each said subinterval a call is placed. A telephone marketing complex management system comprising: the shifting threshold level and means for generating the threshold level for accepting a call. 4. The means for responding to the record holding means comprises (a)
Means for generating a threshold level for shifting a call based on a preselection probability for generating a false alarm to shift and (b) a threshold level for accepting a call based on a preselection probability for generating a false signal to accept The telephone marketing complex management system of claim 3 having both means for generating. 5. The telephone marketing complex management system according to claim 3, wherein said change determining means further comprises line estimation coefficient generating means. 6. Line estimation coefficient generating means for calculating the difference between the actual call load and the estimated call load during each service interval, and an average over the difference during a predetermined number of history periods of each subinterval. 6. The telephone marketing complex management system of claim 5, including means for calculating a squared error. 7. A method of managing a telephone marketing complex comprising an automatic call distributor and means for collecting data about the automatic call distributor during each subinterval of a service interval, the method comprising: each automatic call distributor. Determining the probability in each subinterval that a predetermined performance goal is achieved at the end of the service interval,
Comparing the probability to a threshold level for shifting calls and a threshold level for accepting calls, and determining the number of calls to shift and the number of calls to accept based on the comparison to the threshold level. And storing a record of successes and failures of achievement of the performance goal with corresponding probabilities for each subinterval within the service interval, and calling for each subinterval based on the success or failure record. A method of managing a telephone marketing complex, the method comprising: calculating the threshold level for shifting a call and the threshold level for accepting a call. 8. The telephone marketing complex management method of claim 7, further comprising the step of determining a line estimation factor for each automatic call distributor that can be used to estimate the call load at each automatic call distributor.